It’s clear to see why attribution models are important to a business. I mean, who doesn’t want to know why sales are happening or, how their marketing efforts are driving critical business value. With digital marketing and traditional marketing, attribution is king.
It’s great to know that there is only one true attribution model, since customers only ever interact with one tactic before converting. Right? Isn’t it that simple?
Well unfortunately, no. When talking about attribution there’s myriad approaches that one could take. A popular choice in digital marketing is ‘Last’. What was the most recent tactic a customer interacted with before converting? That model is a great one to start with but it’s easy to see why it doesn’t lend itself well to the complete picture.
That simple reason is, people aren’t that simple. In one case that’s probably a good thing but it does make the concept of attributing a conversion to a tactic more involved.
I’m going to recommend taking a look at Michael Halbrook’s recent post on the core attribution solution offered with the Adobe Insight platform. When reading it, keep in mind that’s just the entry point into the wild world of attribution modeling and analysis. The Adobe Insight technology easily allows you to expand and adapt the modeling as your business changes.
Let me take a few moments more of your time to explain some benefits of using Adobe Insight. With most analysis tools you can access pre-canned aggregated reports. Insight is different. It’s an advanced data analysis platform that builds relationships between event based data to allow for unlimited analysis possibilities.
How does Insight build these relationships?
It all starts with the root of the data model. When a record is processed it’s grouped with other records generated by the same customer (or visitor cookie) and then ordered by time. So instead of taking the information and increasing the daily page view counts or marketing touch counts, that information is actually organized and correlated together.
Figure 1.0 – the generic Insight data model example
It’s like having a file cabinet filled with files based on each customer and their interactions with your business. You can pull out the file on “Doug” and see how he is currently interacting, what he did a week ago, or even how he engaged before making that recent purchase.
Figure 1.1 – the “Doug” Insight data model example
The real benefit of organizing the data in this manner is that it allows you to look at the information from the customer perspective. In the example above, I illustrate the point based on a single customer but Insight provides access to that level of detail as it relates to all your customers and/or segmented groups.
Now that we have all this really well organized information what next?
Insight possesses a rich transformation language that allows one to efficiently modify, adapt and correlate values within the data model. These transformations are what we use to build the attribution models. Here’s a list of a few of the commonly used and more powerful transformations.
The Copy transformation is the simplest but most commonly used. It allows one to copy an input value, field or static string, into a defined output field. The typical application of this transformation is for recasting a field value, correcting/restating a field value, or creating a flag based on a defined conditional like identifying a marketing touch.
The CrossRows transformation is one of the most powerful and complex with the Insight toolbox. It allows one to copy an input from another row within the customer and output it to the current row. That includes prior and future records. The most straight forward example for a use of this transformation would be capturing the channel/name of the last marketing touch prior to a conversion (or micro-conversion).
The FlatFileLookup transformation is very straight-forward and allows the mapping of an input field value to N output fields. The typical application is to provide that friendly name for all the different campaign codes your company is executing on.
The Math transformation is another very straight-forward transformation and enables the use of arithmetic operations on fields. In combination with CrossRows and other transformations, the Math transformation has allowed us to build out some of the more complex attribution models. Such as, Even and Starter/Player/Closer. Where only a fraction of the total conversion value is allocated to the involved marketing touches.
In the definitions above you probably noted the Copy transformation can be used to correct/restate a field value and CrossRows can magically work on future records. What these refer to is how Insight builds its data model for analysis. Insight will read all original raw source data and at that time allow you to apply modifications to field values as they are processed. Subsequently when a new set of data is incorporated within an existing data model, Insight will completely restate how that customer exists within the defined dimensions. So not only do the dimensions fully correlate based on the data model example above they also always reflect the most recent view, or perspective, of each customer.
The real benefits for building an attribution solution within Adobe Insight are the ability to configure the model and get access to the results today, no need to setup and wait, and the inherent configurability of the rich transformation language to extend your understanding of the customer base and how they interact with your business.